Expecting the Unexpected: The Marginal Effect of Unanticipated Terrorist Attacks on Foreign Direct Investment in Israel and Turkey

(1)

1 Master Thesis in Peace and Conflict Studies

Spring 2016

Department of Peace and Conflict Research Uppsala University

Expecting the Unexpected

- The Marginal Effect of Unanticipated Terrorist Attacks on Foreign Direct Investment in Israel and Turkey

ARVID HALLBERG Supevisor: Magnus Öberg

(2)

2 ACKNOWLEDGEMENTS

To everybody who supported me along the way.

I would like to express my gratitude to Flavia Ursa for helping me when the problems with the thesis seemed insurmountable. I would also like to thank Josefin Bergström who allowed me to stay with her when my thesis was completely different and it was unfortunate that in the end that idea did not come into fruition. I would finally like to thank Uppsala University and Magnus Öberg for their support and guidance despite all the minor and major hiccups along the way.

Arvid Hallberg

Uppsala, Sweden 23rd May 2016

(3)

3

Abstract

The paper examines if and to what to extent unanticipated terrorist attacks affect Foreign Direct Investment (FDI) net inflow in Israel and Turkey between 1975 and 2014. The paper utilises two new conceptualisations and operationalisations to determine what makes a year contain unanticipated terrorist attacks and applies them using linear regression on the change in FDI between given years. Using this new operationalisation 10 out of 40 years in Israel and 13 out of 40 years in Turkey were deemed to have experienced unanticipated terrorist attacks. Two models, one controlling for changes in GDP growth, market size, exchange rates and inflation, and one without control variables are used to examine the effect of

unanticipated attacks on FDI. The results indicate that unanticipated terrorist attacks had a statistically significant negative effect on FDI into Israel but no discernible effect in Turkey.

The reason for this discrepancy between the two countries is likely because of the geographic location of terrorist activity in each country, global political pressures, and the type of

terrorist activity.

(4)

4

Tables and Figures

Figure 1: Causal Diagram 17

Figure 2: Scatter Graph of spectrum variable attacks and FDI for Israel 34 Figure 3: Scatter Graph of spectrum variable attacks and FDI for Turkey 34 Table 1: Number of attacks, deaths and wounded in terrorist attacks 28

Table 2: Descriptive statistics for independent variables 29

Table 3: Anticipated and unanticipated attacks Israel 30

Table 4: Anticipated and unanticipated attacks Turkey 30

Table 5: Descriptive statistics for dependent variables 31

Table 6: Correlations Israel 31

Table 7: Correlations Turkey 32

Table 8: Regression table for change in FDI in Israel and binary independent variable 35 Table 9: Regression table for change in FDI in Israel and Spectrum independent variable 35 Table 10: Regression table for change in FDI in Israel and Spectrum unanticipated deaths independent

variable 35

Table 11: Regression table for change in FDI in Israel and binary unanticipated deaths independent

variable 36

Table 12: Regression table for change in FDI in Israel using Spectrum variable attack and control

variables 36

Table 13: Regression table for change in FDI in Israel using binary attack and control variables 37 Table 14: Regression table for change in FDI in Turkey and Spectrum independent variable 38 Table 15: Regression table for change in FDI in Turkey and binary attacks independent variable 38 Table 16: Regression table for change in FDI in Turkey and Spectrum independent variable with

controls 38

Table 17: Regression table for change in FDI in Turkey, binary attacks independent variable with

controls 39

Appendices

Appendix 1: Descriptive statistics of FDI data prior to calculating yearly changes 56

(6)

6

Introduction

Acts of terrorism have a tremendous cost in lives, injuries, material damage, fear and paranoia. Terrorists strike far and wide; from the arrondissements of Paris to the centre of Baghdad they make life less secure for millions of people. Moreover, there are other costs of terrorism, businesses may consider the risks of working in some areas too great, jobs that could have helped people put their children in school or pay for medical services are lost.

Foreign Direct Investment (FDI) contributes to economic growth all over the world and the more globalised the world becomes the more FDI can help countries out of economic turmoil and into sustainable economic growth. Terrorism is calculated to cost countries billions in lost FDI and unrealised economic growths, in 2014 it is estimated that the global economic cost of terrorism was $52.9 billion (Institute for Economics and Peace 2015) as risks

associated with terrorism lower the return on investment and make investors look elsewhere or scrap projects all together. Nevertheless, FDI still flows into countries affected by

terrorism. This phenomenon indicates that some degree of terrorism is acceptable to the risk prone investor.

The relationship between political instability and FDI has been studied at length. Terrorism and FDI have been looked at less and no consensus on whether or not a relationship exists has yet been reached. On the one hand, a significant negative impact has been indicated by Blomberg and Moody (2007) and Alomar and el-Sakka (2011) and Abadie and Gardeazabal (2008) corroborated their findings in large-n studies. On the other hand, many scholars found no significant effect of political stability and terrorism on FDI (for instance Globerman and Shapiro, 2003; Loree and Guisinger, 1995). Li (2006) theorises that the reason for this discrepancy is that all terrorist attacks are not equal. Some terrorist attacks are expected if an investor enters a certain market. His hypothesis is that only unanticipated terrorist attacks have a significant effect on FDI. However, Li only looks at transnational terrorism and ultimately fails to find support for his theory. The aim of this study is to test Li’s hypothesis with a more comprehensive operationalisation of anticipated and unanticipated terrorism.

This paper draws on Li’s work but operationalises anticipated and unanticipated terrorist attacks differently and includes domestic terrorism in the model. This new way of categorising anticipated and unanticipated terrorist attacks is unique and my main contribution to the field of terrorism studies. The primary question asked is: Do foreign investors treat all terrorist attacks the same - contributing to a general feeling of unease and

(7)

7 risk - or do markets price the risk of terrorism into their model leading to a higher marginal effect of unanticipated terrorism as compared to anticipated attacks?

My theory is based on economic theory whereby an investor decides to invest in a foreign market when the expected return on investment is larger than it would be elsewhere, thereby maximising profits. The investor will factor in the expected costs associated to terrorism into their financial model in such a way that the effects of a certain number of terrorist attacks are nullified, at least financially speaking. The attacks that fall outside of this model can be categorised as unanticipated and I hypothesise that they have a significantly larger negative impact on FDI than the anticipated attacks.

In order to test this hypothesis I look at all terrorist attacks in Israel and Turkey between 1975 and 2014 and look at the effects of anticipated and unanticipated terrorist attacks on FDI.

First I develop a new model in order to be able to adequately categorise attacks. Second, I make a regression model to test the effect of anticipated and unanticipated attacks on FDI net inflow the following year.

The results are mixed, with my theoretical framework finding significant support in Israel but not in Turkey. The ambiguities of the results are likely a consequence of an insufficient sample size and more research needs to be conducted to test the results further.

The study is structured as follows: first, a detailed literature review is conducted and a research gap is identified. Second, a theoretical framework is established against which the empirical results are weighed. Third, my research design is presented including the

operationalisation of the key variables and concepts as well as a discussion on the cases under investigation. Fourth, the results are presented. Fifth, the results are critically discussed and alternative explanations put forward based on the limitations of the study. Finally, the summary and conclusion recapitulate the findings and the restrictions of the study and offers implications for future research.

Background and Literature Review

It is safe to say that violence and war has an effect on economic growth. However,

interestingly, the literature on the effect of war differs both empirically and theoretically on whether that effect is positive or negative. The two primary theoretical schools are the so called “War Renewal” and “War Ruin” schools (Kang and Meernik 2005).

(8)

8 The first relies on Keynesian economic theory and is based on Benoit’s hypothesis (Benoit, 1973; 1978) which states that wars have a positive effect on economic growth and

development. The theory suggests that military expenditure should be seen as expansionary fiscal policy and consequently stimulating to the economy, increasing aggregate demand and contributing with positive externalities such as technological progress, improved

infrastructure, and employment opportunities (Ganegodage and Rambaldi 2012).

The second argues that the opportunity cost of war does not make up for the economic benefits derived from wartime expenditure. Investments in other important areas such as health care and education are crowded-out, foreign direct investment decreases and, private investment is reduced and talented individuals seek opportunities elsewhere creating a brain drain effect on the economy (Collier 1995; Ganegodage and Rambaldi 2012; Imai and Weinstein 2000; Schonfelder 2005).

In the same vein as the two schools of war, the literature on political instability, political risk and terrorism, and their effects on the economy is ambiguous and inconsistent. Contrary to the schools of war however, almost all terrorism scholars agree that terrorism has a

detrimental effect on the economy. Using the two schools of war as a theoretical backdrop this paper deals with unanticipated terrorism’s effect on FDI. As previous literature on the effects of unanticipated terrorism, or indeed, terrorism in general on FDI is relatively superficial the literature review will take a broader view and include tangential but

theoretically similar concepts such as political instability and political risk. As terrorism lacks a universally accepted definition some authors include violent social unrest and political violence under a different umbrella (for instance Shapiro 2003). As there is valuable insight to draw from their research, and the fact that the causal relationship between unanticipated terrorism and FDI, and unanticipated political violence and FDI is very similar they are included in this paper.

Early studies found empirical irregularities regarding the effects of political instability and FDI (Kobrin 1979). Schneider and Frey (1985) find that FDI net inflow is negatively affected by political instability and Nigh (1985) finds through an analysis of 24 countries over 21 years that conflict has a negative effect on FDI flows by US firms. However, Fatehi- Sedeh and Safizadeh (1989) do not find evidence of statistical association between political

(9)

9 stability and FDI net inflow. Loree and Guisinger (1995) find in an analysis of 36 countries in 1977 and 1982 that stability promotes FDI in 1982 but not in 1977. For FDI into OPEC countries, Olibe and Crumbley (1997) do not find evidence that indices of political risk influenced US FDI flows in 10 out of 13 cases. Li and Resnick (2003) also do not find any link between political instability and FDI inflows, but they do find a link between regime durability and FDI. In the same fashion, Sethi, Guisinger, Phelan, and Berg (2003) find that political instability does not influence U.S. FDI flows to 28 countries from 1981 to 2000.

Looking at Pakistan between 2005 and 2014 Shakeel and Shah (2015) found a negative relationship between GDP and suicide attacks and drone strikes, whereas no negative relationship was found between FDI and suicide attacks and drone strikes. However, Globerman and Shapiro (2003) find that although an index of political instability and violence, including but not limited to armed conflict, social unrest, and terrorism does not influence the probability whether a country receives any FDI inflow, it does reduce the amount of FDI inflow a country receives.

Two seminal studies on the link between terrorism and FDI are Busse and Hefeker (2007), and Blomberg and Mody (2007). The former found that governmental stability, religious tensions, and democratic accountability are the three primary political indicators that impact the net inflow of FDI. The latter looks at the impact of violence on trade and bilateral FDI flows between 12 source countries and 43 host countries. Their results indicate that violence has a significant negative impact on FDI and trade, and for developing countries it induces a shift of FDI from horizontal to vertical¹.

Other scholarship paints a more nuanced but nevertheless still inconsistent picture. Power and Choi (2012) find that terrorism that targets multinational investments in developing states negatively affects FDI net inflow to those states but terrorist attacks that do not target businesses have no statistically significant effect on FDI. Enders, Sachsida and Sandler (2006) also suggest that it depends on the economic development of the country and find that terrorist incidents have a significant effect on US FDI in OECD markets but the effect

disappears in non-OECD economies. Others disagree and find that terrorism has an

unequivocally negative effect on FDI. For example, Alomar and el-Sakka (2011) and Abadie

1 Horizontal FDI refers to a firm locating production abroad to be closer to customers and cutting down on transportation costs, i.e. horizontal FDI is trade with goods to parties outside the firm. Vertical FDI refers to trade to parties within the firm and is used when some factors of production are cheaper than they would be in the home country. (Ramodo, Rappoport, and Ruhl 2014)

(10)

10 and Gardeazabal (2008) using samples of 136 and 186 countries respectively find a negative correlation between terrorism and direct investments. An increase in terrorist risk in the afflicted country of one standard deviation is concluded to cause a five percent decrease in foreign investment flows as a share of GDP. This is supported by Lutz and Lutz (2006) who suggest terrorism caused a substantial decrease in FDI for Latin American countries and by Mehmood (2013) who concludes that cumulatively terrorism has cost Pakistan 1% of GDP per capita a year for the 14-year period 2000-2013.

Other evidence such as responses to surveys, questionnaires and interviews suggest that political risk and instability are major questions when investors decide on investment decisions (see for example: Bass, McGreggor, & Walters, 1977; Porcano 1993). In a study using a sample of 25 countries and spanning from 1995 to 2010 Stanisic (2013) examines how investors distribute their investments between host countries and finds that terrorism plays a significant part. His results suggest that terrorism is a disincentive to investment and causes investors to look elsewhere, at less risky economies, for their capital (Ibid: 25).

We know that terrorist attacks affect the economy both directly and indirectly. However, as society has grown more used to the risks associated with terrorism firms may have learned to price in the risk and uncertainty into the initial investment. Thus, different terrorist attacks may have different effects on investments. On this subject, Li (2006) finds that “anticipated terrorist incidents do not produce any statistically significant effect on the chance that a country is chosen as an investment location or the amount of FDI it receives.” Which is logical as anticipated attacks would be included in the cost calculations. However, “Contrary to our expectation, the unanticipated terrorist incidents, despite their unexpectedness, do not generate any changes in investor behaviours, either in terms of the investment location choice or the decision over investment amount.” (Li 2006: 250-51). Li does not provide an adequate answer as to why unanticipated terrorist incidents do not affect investments but urges further research to investigate further. This peculiar result has not been addressed in the terrorism- FDI literature nor in any other econometric studies.

Thus, the purpose of my study is to draw from Li’s claims that anticipated and unanticipated terrorist incidents have no statistically significant effect on FDI net inflows into a country.

With a new operationalisation the marginal effect of unanticipated attacks can be presented

(11)

11 and thereby improve our comprehension of how terrorism affects FDI net inflow.

Understanding what kind of terrorist attacks have a significant impact on financial confidence is important for identifying what our immediate response to an attack should be as well as how we should prioritise our efforts in counter terrorism. Further, a better awareness of the inconsistency between whether investors do take terrorism into account or not has important implications. Foremost, it has important theoretical implications vis-à-vis our image of the relationship between politics and international production. If terrorist attacks have no significant effect then investors can allocate their investments more efficiently than they do today and consequently increase their profits and the overall growth (Li 2006). On the contrary, if investors do in fact consider the importance of political instability and terrorism the econometric findings suggest a flaw in our understanding of the relationship between terrorism and international production (Ibid).

Second, the question is important for understanding the effect of FDI on political violence.

Authors (e.g., Gartzke, Li, and Boehmer 2001; Gartzke and Li, 2003) have found a negative relationship between FDI and civil conflict in developing countries as FDI leads to a larger pool of available jobs and resources. While in more developed states Rothgeb (1990) find that they intensify class struggles as the added resources from FDI are not sufficient to satisfy the needs of the population and are instead taken advantage of by a privileged elite. Further, FDI inflows appear to have a stabilising effect on transnational terrorist attacks by promoting economic development (Li and Schaub 2004). All of these papers consider it axiomatic that violence affects investment flows.

Hence, based on previous research inconsistencies and surprising findings I present the following research question:

Do foreign investors treat all terrorist attacks the same - contributing to a general feeling of unease and risk - or do markets price the risk of terrorism into their model leading to a higher marginal effect of unanticipated terrorism as compared to anticipated attacks?

Theoretical Framework

(12)

12 Terrorism has a myriad of macroeconomic effects that will be discussed in detail below. This section is structured into three parts. First, it will examine why investors choose FDI as an avenue and the attributes of international production. Second, the ex ante effects of terrorism on FDI will be discussed. Third, the ex post effects of an attack will be theoretically

examined.

Why FDI - or the features of international production

In order to understand the effects of terrorism on FDI it is important to understand the theories behind it. Most economists agree that FDI is a vital element for economic

development, especially in developing countries (Denisia 2010). FDI means higher exports, access to previously unavailable markets and currencies, as well as a substitute source of financing (Ibid. Goldberg 2004). Additionally, there is evidence that FDI reduces

unemployment and improves the competitiveness of domestic companies (see for example Findlay 1978; Caves 1996; Borenzstein and de Grigorio 1998). Others have argued that FDI may crowd out national investments and that the effects of FDI are mostly negative (for example Hanson 2001; Gorg and Greenwood 2002) or that it depends on the sector of the economy as there appears to be very limited benefits in the agricultural and mining sectors (Hirschman 1958; Khaliq and Noy 2007).

The focus of this paper – FDI net inflow – refers to investments made into another country by non-residents. FDI normally means the investor acquires a significant degree of management influence (normally 10% or more of voting stock) (Li 2006: 234). Additionally, direct

investments include “investment associated with that relationship, including investment in indirectly influenced or controlled enterprises, investment in fellow enterprises (enterprises controlled by the same direct investor), debt (except selected debt), and reverse investment.”

(World Bank Undated)

The literature on FDI provides conflicting theoretical explanations of why firms engage in international production. Some make the argument that FDI is a result of market imperfection and that the investing firm uses FDI in an attempt to reach a monopoly situation (e.g. Hymer 1976). Others see FDI as a tool to pressure and control foreign markets, as opposed to entering the market in other ways such as trade (see Buckley and Casson 1976; Rugman

(13)

13 1981). Another school of thought (Vernon 1966, 1971) consider FDI as a response to

technological maturity of a firm’s product and growing demand in a foreign market.

However, the most comprehensive theory is known as the eclectic paradigm (Dunning 1980, 1988, 1993) and it ties the previous explanations together forming an OLI framework, meaning firms seek to exploit three types of advantages. O – The advantage of ownership over both tangible and intangible assets. Ownership can be divided into three distinct advantages: monopoly advantages such as privileged access to markets, patents or trademarks; technological advantages vis-à-vis local and international competitors; and economies of scale. L – Location specific advantages such as natural resources, tax or other government policies, as well as access to talented personnel. I – internalisation advantages from control over cross-border production – meaning it must be profitable to use the O and L advantages in conjunction with factors outside the home country (Denisia 2010: 108). In short, a firm decides to invest directly abroad when the location and direct ownership

provides for profitable production abroad and where direct control is preferred to outsourcing or trade.

Building on these attributes of international production, Li (2006) adds two points that are important to consider with regards to how terrorism affects FDI: cross-border jurisdiction and a long time-horizon. Li makes the argument that host policies on “expropriation, exchange control, breach of contract, repatriation of profits, voluntary divestment, performance requirement, taxation,” are of critical importance and if these policies can be influenced by terrorism then terrorism will be taken into account when investors make decisions (Li 2006:

235-36).

Ex Ante effects

Li’s second point, a long time-horizon, combines two fundamental facts of FDI and

international production. First, the investment itself is a barrier to exit as the investor cannot go back on his investment without incurring some costs (Rivoli and Salorio 1996). Second, this barrier means the investor must be anticipating the political and economic development of the country or region including potential political violence and terrorism (Li 2006). In other words, working on a long time-horizon the investor will calculate the projected expected profit based on both OLI advantages and risks. This evaluation then becomes the basis on which the firm makes an investment decision. On an empirical level, survey answers

(14)

14 suggest that executives take political instability into account political instability when making investment decisions (see for example Bass, McGreggor and Walters 1977; Porcano 1993).

Thus, an otherwise lucrative investment can become too expensive if the investor anticipates the risk of terrorism to be too high to offset the projected profits the investment. Additionally, a change in the risk of terrorism will affect the projected earnings and value of current

investments and consequently deter future inflow, limiting growth and even prompting divestment away from the country or region.

It is important to remember that these calculations take place regardless of whether a terrorist attack happens or not and potentially unfavourable consequences and government responses to an attack would be included. Therefore, if an anticipated terrorist attack happens it should have little to no effect on FDI inflows ex post assuming the risk-adjusted returns are the same before and after an attack.

Ex post effects

Terrorism has a myriad of macroeconomic effects ex post; production, equity markets, capital account, transaction costs, government spending, inflation are all impacted in various ways.

Indirectly it also affects trade through tighter border controls and other barriers (McKenna 2006). Multiple works have been published on the effects of terrorism on the economy in general and on FDI inflow in particular.

One of the main indirect effects of terrorism is the increase in international trade transaction costs. Importantly, the market response appears to be governed more by the policy reactions to attacks than to the attacks themselves (Chen and Siems 2004). Blomberg et al. (2004) find that a terrorist attack is associated with a change of spending, from investment to government expenditure. This necessitates increased borrowing from both other governments and

institutional investors. Gupta, Clements and Inchauste (2004) conclude that, from a monetary policy perspective, this bout of borrowing negatively affects less developed states more than developed economies with less volatile currencies as developing states are forced to prioritise the military and security sector over investments in economically viable sectors. Confronting the terrorist threat requires using their foreign reserves, increasing inflation whereas for OECD nations issuing debt does not have many serious long-term repercussions (Gupta,

(15)

15 Clements and Inchauste 2004). However, over time inflation declines as market equilibria are re-established. Other policies that directly influence international trade include counter- terrorism measures such as tighter border controls with close inspections of people, vehicles, and goods as well as tightened immigration laws (Brück and Wikström 2004). Some

governments may choose to react unfavourably towards the investors by for instance

expropriation or through unanticipated tax hikes. Walkenhorst and Dihel (2002) estimate the scale of the cost increase due to terrorism vary between 0,5% and 3% ad valorem. This cost increase has a direct effect on trade flow and a doubling of terror incidents reduces bilateral trade by 4% (Nitsch and Schumacher 2004).

These general claims of terrorism’s effect on FDI are backed up by case studies, for example Ullah and Rahman (2014) shows that terrorism had an adverse effect on Pakistan’s economy due to diverting government expenditure towards counter-terrorism as well as causing other more direct costs such as loss of life, destroyed infrastructure, and trade restrictions.

Terrorist attacks poses a significant risk to production, particularly oil producing states are vulnerable through attacks on oil infrastructure such as pipelines and other distribution mechanisms (Nitsch and Schumacher 2004). Uncertain production levels inhibits the central government’s ability to estimate taxes on actors in the petroleum sector and consequently contributes to higher uncertainty levels for tax revenues. In addition, terrorist attacks raises the cost of doing business as firms are forced to invest in more security, pay workers risk premiums and increases uncertainty in general (Mazzarella 2001).

Due to the speed of information equity markets are also quickly affected by terrorism. The addition of the uncertainty introduced by an unanticipated terrorist attack may force investors to re-evaluate their investments and with a lack of information can quickly spread investor panic affecting the country’s equity market (McKenna 2006). As investors get access to more information they will make assessments about their firms’ and investments’ resilience to potential political, societal and economic changes. Considering that stock markets indices and indeed currency rates are supposed to be representative of the combined equity and currency portfolios, the market response is a consequence of the aggregate portfolios in that market (McKenna 2006: 4). Chen and Siems (2004) suggest that the degree of exposure to terrorist risk depends on the financial sector’s supply of liquidity; higher liquidity promotes a buffer and acts as a stabiliser against fluctuations in investor confidence. In a paper examining the

(16)

16 effects of terrorism on Israel’s economy Eldor and Melnick (2004) found that despite the ongoing terror attacks against the country, 639 attacks between 1990 and 2003, the foreign exchange market was not affected. However, the stock market appears to have internalised future attacks by reduced expectations on future profits. In other words, the stock market acted as foreign investors would, anticipating future attacks, and the price of terror was reflected by lower projected future profits.

What all these macroeconomic effects have in common is that they increase the level of uncertainty for an investor. Investment uncertainty can be either endogenous or exogenous, endogenous uncertainty can be resolved through experiential learning (Rivoli and Salario 1996). Anticipated terrorism, already a part of the investor’s calculations is a known unknown and an endogenous uncertainty. However, unanticipated terrorism contributes to uncertainty that is exogenous to the investment itself forcing investors to re-evaluate their models and incorporating additional information such as government responses and new risk- adjusted returns. Note that it is not how big the effect of the terrorist attack that matters but how big the unanticipated effect of that attack becomes. The ex ante and ex post effects are inversely correlated (Li 2006). A large anticipated attack would, according to the theory, be internalised and have a small ex post effect whereas an unanticipated small attack would have a larger effect due to the fallout not being already internalised (Ibid.).

Causal Argument

Thus, based on the theory outlined above, a number of short assumptions can be drawn. First, investors choose to invest in a foreign market because the benefits outweigh the costs

weighing the projected profits against the projected risks. Second, the effects of terrorism, though unknown, can be included in the projected risks of an investment. Third, anticipated terrorist attacks and their effects would be mostly included in such model whereas

unanticipated terrorist attacks would not. Therefore, unanticipated terrorist attacks should have a bigger impact on FDI net inflow into a country or region as the costs related to terrorism as well as the increased uncertainty makes investments there more expensive.

Causal Diagram

(17)

17 A simple causal diagram can help illustrate this theory further. The x-axis shows the return on investment. The y-axis shows the costs of doing business in a foreign country. The three lines illustrate that anticipated terrorist attacks change the constant costs but otherwise remains parallel to investments in countries without terrorism. The unanticipated attacks raise the costs significantly but does not change the constant as they were not accounted for prior to making the investment. The point of this diagram is to illustrate the theoretical framework, not to make any actual illustration of the ratio between terrorism and costs, hence the slopes of the lines should not be understood as actual predictions but merely a sketch of the

relationship.

Causal Diagram Figue 1.

Hypotheses

Based on this theory, a simple hypothesis can be established. Despite the arguments of the

“War Renewal” school it appears highly unlikely that terrorism, whether anticipated or unanticipated, has a positive effect on FDI. Any positive relationship is certainly due to weaknesses in the operationalisation or insufficient data. For this reason this study uses one-

(18)

18 tailed tests and one-tailed hypotheses to test the relationship and uses the cases of Israel and Turkey.

H: Terrorist attacks will affect the FDI net inflow into a country differently depending on whether the attacks could be seen as anticipated or unanticipated.

H0: null hypothesis: There is no relationship between anticipated or unanticipated terrorist attacks and FDI net inflow.

H1: There is a significant negative relationship between unanticipated terrorist attacks and FDI net inflow

Definitions

Before going into detail about my research design a few key concepts need to be theoretically defined. This section focuses on descriptive definitions, the operational definitions used in the actual model are found in the Research Design section below.

My dependent variable, FDI net inflow, is a category of cross-border direct investment and refers to the “capital transactions' credits less debits between direct investors and their foreign affiliates” (World Bank) that flow into and out of the host economy. Net decreases in assets and increases in liabilities are documented as credits and vice versa (Ibid). When FDI net inflow is negative at least one of the components is negative and not offset by the other components. These years are important as they are instances of reverse investment or disinvestment (Ibid). Direct investment is also associated with a foreign resident acquiring a significant influence on the management of a host-country enterprise. 10% or more of the ordinary shares voting stock is the criterion used by the World Bank for determining a direct investment relationship. In order to better capture the effects of unanticipated terror this paper measure the change in FDI from the previous year instead of FDI in absolute terms. This change is measured in percentage.

Methodology

(19)

19

Cases

Previous research into terrorism’s effect on FDI mainly consists of Large-N studies on 100+

countries (for example; Abadie and Garderazabal 2005), or in depth case studies of one country (Abadie and Gardeazabal 2008; Elders and Sandler 1996). Only one study (Li 2006) looked at anticipated and unanticipated terrorist attacks and only in a large-n study.

Counterintuitively, Li did not find an effect of unanticipated attacks on FDI. Li only looked at transnational terrorism which is likely to have skewed his results as many terrorist attacks are perpetrated by domestic groups.

In this study, I look at the effects of unanticipated terrorism on two Mediterranean countries;

Israel and Turkey. Israel has since its inception been faced with both conventional threats from neighbouring armed forces and unconventional threats from mainly Palestinian terrorist organisations such as the Palestine Liberation Organisation (PLO), the al-Aqsa Martyrs Brigade, Democratic Front for the Liberation of Palestine (DFLP) and Hamas. Hezbollah in Lebanon has also showed themselves capable of launching attacks into the North of Israel.

During the time frame for this paper, a number of notable peace treaties were signed as well.

The three most important are arguably the 1979 treaty with Egypt, the 1993 Oslo Accords with the PLO, and the 1994 treaty with Jordan. Israel’s peace with Egypt in 1979 may have been offset by the breakdown of relations with its previous most significant regional ally Iran following the Islamic Revolution. (Arian 2011).

Turkey has been plagued by several terrorist organisations including the Kurdistan Workers’

Party (PKK), the Turkish Workers’ and Peasants’ Liberation Army (TIKKO) and more recently spillover attacks from ISIL/Daesh and Kurdish groups in Syria. Turkey has had a history of military intrusions into politics and experienced a number of coups during the time frame. First in 1980 with a “regular” institutional coup, but arguably also in 1997 when the military eventually ousted the existing coalition government by including the pro-Islamist Welfare Party (Eder 2011). The coup in 1980 also led to a worsening position for the

country’s Kurds and between 1984 and 1999 Abdullah Öcalan’s PKK fought using terrorism tactics against the Turkish government (Ibid: 738). A state of emergency was declared and not lifted until 2002. Finally, an important aspect to consider when looking at Turkey is the economic divide between East and West. The rate of extreme poverty is five times higher in the southeast than the country average. The western part of the country is more integrated in

(20)

20 the world economy with high levels of tourism, investment, better infrastructure and accounts for a much larger share of the country’s GDP (Ibid: 734).

I focus on these two countries for three main reasons. First, tourism which has been shown to be the industry hit very hard by terrorism (Drakos and Kutan 2003) is an important sector in both Israel and Turkey contributing with between 7 and 9% and between 10 and 14% of GDP respectively since 2000 (World Travel and Tourism Council Data 2016). Second, both

countries have seen peaks and troughs in terrorism activity with periods of calms and uprisings throughout the time frame, this makes it likely that there are numerous years with unanticipated terror attacks. The periods of calm as well as periods of intense fighting also provide a good opportunity to see how FDI net inflow changed during the peaks and troughs of terror activity. Third, as centres of terrorism activity for a number of groups as mentioned above these countries offer an opportunity to look at the effects of large-scale terrorism on a national level. Both Turkey and Israel have had concerns with terrorism meaning that it is an issue that any investor would have to include in an investment risk analysis. The domestic situation in both countries means that there are very few years in the dataset without any terror attacks at all. As such, these cases can act as probes for future research by highlighting aspects that make unanticipated terrorism affects the net inflow of FDI.

Using the two cases of Turkey and Israel I will be able to both run my own statistical analysis on anticipated and unanticipated attacks, and look more in depth at the reasons for its

statistical significance or insignificance. Looking at only two cases allows me to complement the statistical analysis with a more qualitative discussion taking into consideration things such as what was the cause of increased terrorism in certain years, geographic location of terrorist attacks, the industrial and economic makeup of the countries, and other political development that may impact FDI. As the research on anticipated and unanticipated terrorist attacks is very sparse I have chosen two representative cases, in the sense that issues and results found in Israel and Turkey should be translatable to other countries beleaguered by terrorist activity, in order to provide a solid foundation for continued research and theory building.

Using representative cases, or plausibility probes, allows me to develop my theory in an exploratory approach and refine the operationalisation of my independent variable. As my independent variable is meant to capture the financial modelling used by investors, who undoubtedly are using more complex models for determining the likely number and effects of

(21)

21 terror attacks, this study should be understood as a jumping off point - an exploratory analysis - to generate ideas on which to build more elaborate models more than a true judgment of the theory itself.

Material and sources

The data on FDI and economic control variables were taken from the World Bank. The data on terrorism was taken from the Global Terrorism Database (GTD).

The World Bank provides data on a number of economic and social variables and divided up by country and year. For monetary amounts the World Bank converts the local currency to 2005 dollar amounts. This means it is possible to compare cases over time and as all of my economic variables come from the World Bank there should be no methodological flaws in comparing them. Another option was using data from the countries’ respective central banks, however the World Bank uses data from the International Monetary Fund supplemented by data from the United Nations Conference on Trade and Development as well as official national sources and the numbers provided by the World Bank are identical to the numbers provided by the central banks. The two major benefits of using the World Bank data are the availability and the fact that it goes back further in time than the data easily available from official national sources.

The GTD dataset is more complicated and has a number of problems that must be discussed.

The GTD has been managed by a number of organisations with several phases of data collection (GTD Codebook 2015). This means that the methodology used has changed over time, for example, when START took over data collection in 2012 there was a dramatic increase in number of terrorist attacks compared to 2011. According to the GTD “this increase likely reflects recent patterns of terrorism, it is also partly a result of the improved efficiency of the data collection process” (Ibid: 8). The internet, improved archives, and more complex algorithms have made data collection on terrorist attacks much simpler and more efficient, thus it is likely that more attacks are missing from the dataset in the early years than more recently. Indeed, the GTD themselves state that they “continue to work to supplement the GTD “legacy” data back to 1970 to further improve its completeness” (Ibid: 8). The consequences for my research are a potential bias towards a smaller effect of terrorism as a year with unanticipated terrorist attacks may be coded as having none. Further, when the

(22)

22 GTD transferred ownership in 1997 the data on 1993 was lost. It has been recreated and is included as an appendix in the dataset but a large number of variables missing. As this paper is mainly concerned with the number of attacks, number of people killed, number of people wounded, and the fact that 1993 was a comparatively violent year in both Turkey and Israel the year is still included in my dataset. Again, this may bias my results, especially in the case of Turkey where there were likely more attacks than reported but according to the data there were almost exactly the anticipated number of attacks.

The GTD uses its own definition of a terrorist attacks and defines it as “the threatened or actual use of illegal force and violence by a non‐state actor to attain a political, economic, religious, or social goal through fear, coercion, or intimidation” (Ibid: 8). First, this excludes state actors but includes paramilitary organisations. In the case of Israel it poses several questions on how attacks were classified during the wars against Lebanon in 1982 and 2006.

Attacks launched by Hezbollah and PLO would be classified as terrorist incidents whereas attacks by the Syrian Army would not. It is unclear whether joint attacks by Syrian and Hezbollah forces would be classified as terrorism or ignored as an act of war.

Moreover, as terrorism lacks a comprehensive definition there is bound to be an overlap between acts of terror and other forms of violence. The GTD includes instances of terrorism when there is doubt whether the attacks were an act of terror meaning that there could be a higher number of terrorist attacks reported for certain years than took place in reality. This again is problematic as years coded with unanticipated terrorist attacks may in reality not have had any and therefore the expected effect on FDI is absent. It is worth noting that the GTD has a variable for these acts but only from 1997 and onwards, for reasons of consistency I have omitted the variable from my statistical analysis.

Another issue with the GTD coding is that they code numerous terrorist attacks happening in conjunction as separate incidents. Or as stated by the GTD: “If the information available for a complex event does not specify a time lag between, or the exact locations of, multiple

terrorist activities, the event is a single incident. If any discontinuity in time or space is noted, the event is comprised of multiple incidents.” This means that a coordinated attack by four suicide bombers at the same time in the same city is still coded as four attacks. Whether or not investors see it as one or four attacks is unknown and it may therefore again result in

(23)

23 years coded as having unanticipated attacks when in the mind of investors the particular year followed the risk model they developed prior to making the investment.

Time frame

This essay uses the data from 1975 to 2014. The reason for this is simple; the GTD dataset starts in 1970 and for operational reasons the earliest possible start is in 1975. Theoretically, the time period is not of great importance although potentially financial models are more advanced and information more readily available in the 2000s than in the 70s and 80s. Issues pertaining the terrorism data were discussed above, but for the economic variables the time period may introduce confounding variable such as recessions and periods of high growth when investors may disinvest despite a lack of terrorist attacks or invest despite more terrorist attacks than their model suggests as the rapid economic growth is lucrative enough to accept the higher risk of terrorism.

Operational Definitions

My dependent variable is the change in FDI net inflow per year compared to the previous year and measured in a percentage. In mathematical terms: FDI net inflow for year is equal to (t-(t-1)/t)*100. Mathematically, this is identical to using FDI net inflow and controlling for the previous year.

My independent variable is whether or not attacks in year t can be classified as anticipated or unanticipated and in the latter case how unanticipated they are. As there is no set definition of terrorism and as the effect of a terrorist attack is largely governed by the governmental and societal response to an attack I use three separate indicators that should indicate the severity of terrorism in a given year: namely, number of attacks, number of people killed, and number of people wounded. The extent of damage on property was excluded as there were a lot of missing variables.

For all three variables I calculate a moving average (MA) of the mean of the previous five years, I then calculate the difference between the actual number of attacks in year t and divide it by my MA plus one standard deviation of the past five years. If this variable score is below 1 all attacks in year t are classified as anticipated, if the variable score is above 1 some of the

(24)

24 attacks should be considered unanticipated and the larger the variable score the more

unanticipated attacks took place. I call this variable “numerical binary”. I also use my

numerical binary variable to create a true binary variable where 0 represents the years without unanticipated terrorist attacks and 1 represents years with unanticipated attacks. This variable does not capture the degree of how unanticipated attacks were, for instance is the predicted number of attacks plus one standard deviation were 20 and the actual number of attacks were 21 it is coded thee same way as if the actual number of attacks were 100. Finally, I use a third operationalisation which I call a spectrum variable where I take the actual number of attacks minus the anticipate number of attacks divided by one standard deviation. Again this gives me a situation where a number larger than 1 represents a year with unanticipated attacks and a number below one represents a year without unanticipated attacks. As my independent variables are meant to act as dummy variables for the financial models used by investors it is worthwhile to test different calculations against one another to increase the confidence in my results.

Mathematically my independent variables can be written like this:

1) Numerical Binary y1=x/( μ+ σx)

2) Binary variable

y2=1 if 1) >1 and 0 if 1)<1 3) Spectrum variable y3=(x-μ)/ σx

Where x is the actual number of attacks for a given year, μ is the mean of the previous five years of x and σx is the standard deviation of the previous five years of x. Note that 1) is only used to obtain the binary variable and not presented in the results section. The reason for this is that the variable did not present any additional information that was not clear through the binary variable and the spectrum variable. The same operationalisations are used for binary and spectrum deaths and wounded respectively.

Using the mean plus one standard deviation should capture the years where attacks were unanticipated. A theoretical issue is if you assume that investors are risk averse and already

(25)

25 include one or even two standard deviations in their price models. If this is the case my model is not valid.

Control Variables

A set of control variables are used to strengthen the predictive power of the model. Namely, market size measured in current USD, GDP growth, inflation, and currency strength. All the control variables are measured in change from previous year. A problem with using these control variables is that prior to making an investment the investor would likely anticipate how these values are likely to change for the coming years. A positive growth rate may not be correlated to an increase in FDI if the anticipated growth rate was higher than the actual outcome. Similar to the theory on anticipated and unanticipated attacks these values would be priced into the investment. This is problematic for my model as I use the absolute changes of the control variables instead of the anticipated or unanticipated changes. The reason I have chosen to opt for the absolute changes instead of calculating the anticipated values is to limit the number of assumptions necessary for the model. Additionally, even investors’ models may prove themselves short and as the statistics compare the FDI inflow in year t with terrorism in the same year it is best to assume perfect information for all other variables in order to set the effect of terrorism apart.

The most robust determinant for FDI is market size as bigger markets with greater purchasing power offer investors higher return on their investment (Artige and Nicolini 2005; Jordaan 2004; Demirhan and Masca 2008). Charkrabarti (2001) hypothesises that the reason market size is important is because a larger market can utilise resources more efficiently exploiting economies of scale. Market size is such an important component of FDI that nearly all studies on determinants of FDI has market size as an explanatory variable (Demirhan and Masca 2008: 358).

Market size is not theoretically linked with terrorism. It is, however, linked with economic performance and poverty which in turn is linked with terrorism. The relationship between poverty and terrorism is a matter up for debate with many scholars claiming there is no relationship (e.g. Gassebner and Luechinger 2011; Piazza 2006)whilst others claim the relationship is non-linear (Enders and Hoover 2012; Enders, Hoover and Sandler 2014) with terrorism being more prevalent in middle-income countries compared to rich or poor

(26)

26 countries. Changes in market size is also closely linked with FDI net inflow. For this reason, it is included in my control model to ensure that changes in investor behaviour are not falsely attributed to terrorism when an increased or decreased market size may be the true factor affecting investor behaviour. An expanding and rapidly growing economy should also attract more FDI to the host country. The growth hypothesis states that a faster growing economy provides better opportunities for profit making relative to a slower growing economy

(Charkrabarti 2001; Lim 1983). Many scholars (Lunn 1980; Schneider and Frey 1985; Culem 1988) have found support for this hypothesis stating that growth has a significantly positive effect on FDI.

Economic growth and consequently an increased GDP is intricately linked with poverty. The relationship between poverty and terrorism is a contested debate. Most scholars reject the relationship, claiming that terrorism has more to do with inequality, lack of political voice or freedoms, transition towards democracy, and other geographic factors (Abadie 2004).

Equally, Gassebner and Luechinger (2011) do not find any causal relationship between per capita GDP and terrorism in their results from 13.4 million different regressions. However, Enders and Hoover (2012) find that by separating domestic and transnational terrorism poverty has a distinct effect on each type of terrorism. Contrary to popular research, they find that poverty has a strong nonlinear effect on both domestic and transnational terrorism. The nonlinearity of the effect means that regular linear regression for the relationship between GDP per capita and terrorism does not uncover the effect (Enders and Hoover 2012: 272).

Additionally, as the main perpetrators of violent, systemic terrorism has changed from being mainly leftist groups in the 70s and 80s attacking targets in rich countries, to mainly religious fundamentalists attacking targets of opportunity and countrymen in poorer countries the causal link between GDP and terrorism is likely to have changed over the time period used in this paper (Enders, Hoover and Sandler 2014). Similarly to market size, the nonlinearity of the relationship makes the potential confounder difficult to control for and for this reason this paper will present the results in two models, one controlling for confounders and another stripped of all control variables.

High or increasing inflation is anticipated to have a negative effect on FDI as it alters the net returns of the investor and changes optimal investment decisions (Sayek 2009). On the one hand, inflation lowers the purchasing powers of individuals and consequently gives the investor a lower return of investment. On the other, some degree of inflation is necessary for

(27)

27 the economy. I use the change in inflation mainly to control for inflation shocks which could lead to increased terror as inflation lags and unemployment increases² or divestment if inflation spikes.

Currency stability is another component that has a strong effect on FDI. Currency stability should not be understood as a variable that affects terrorism but rather as a middle variable affected by terrorism and in turn affecting FDI inflow. Researchers in Pakistan (Qaiser, Sohail, Liaqat, and Mumtaz 2012) using data between 2007 and 2010 show that an increase in terrorism during this period depreciated the Pakistani currency significantly. Tavor (2011) examines the effect of terrorism on the capital market in Israel and finds that the location, intensity, and the response will determine the effect or lack thereof of terrorism on the capital markets.

In this paper currency stability is defined as the rate between local currency units and constant 2005 US dollars. The exchange rate has an effect on FDI both in terms of the strength of the currency and its volatility (Goldberg 2006). A depreciating currency relative to the currency of the investor means wages and production costs decrease in the host country. All else being equal, a depreciating currency leads to “locational advantage”, a higher rate of return and should have a positive effect on attracting FDI into the country (Ibid:

1). The volatility of the local currency also affects the FDI inflow into a country. The

production flexibility theory states that “more volatility is associated with more FDI ex ante, and more potential for excess capacity and production shifting ex post, after exchange rates are observed” (Ibid: 4). A counter argument is found in the risk averseness theory which states that the uncertainty associated with a volatile currency adds a risk to the potential returns on investments (Ibid: 5). The two theories offer different directional predictions on the effect of currency volatility, however both maintain that the exchange rate has a

significant effect on FDI³ This effect is a potential confounder for my results and should thus be controlled for.

2 Unemployment, and especially youth unemployment, can lead to participation in terror groups (and indeed gangs and other organised violence) for a variety of reasons such as pecuniary, status, norms of fairness, ethnic bonds etc (Cramer 2010). For further reading see for example Caruso and Gavrilova 2012.

3 For further reading on the effects of exchange rates on FDI see for example Aizenman (1992), Devereux and Engel (1998), and Russ (2007).

(28)

28

Method, statistical analysis and assumptions

A visual inspection of the Q-Q plots and histograms showed that my dependent variable is normally distributed. As the dependent variable is quantitative, normally distributed and given the fact that there are no latent variables I choose to estimate the effect of unanticipated terrorist attacks on FDI using OLS regression. OLS regression is also the most useful for finding the estimated effect of unanticipated terrorism on FDI net inflow. In addition to linear regression I will use descriptive statistics and correlations to simplify my data and results.

The linear regression tests are divided into binary and spectrum attacks, deaths and wounded and supplemented with tests where the control variables are included. As discussed above, the theoretical link between the control variables and terrorism is meagre but present. Using both models will first ascertain the effect of unanticipated attacks and then test that finding by including control variables. By testing both with and without control variables the paper overcomes the pitfall of “trashcan regressions” (Aachen 2005).

In the discussion section the statistical results will be complemented by a more in depth qualitative look where I will observe the years with unanticipated attacks to see if there was anything else out of the ordinary apart from increased terrorism. In effect, the years with unanticipated attacks will act as a foundation for a short qualitative analysis of the results.

This approach means I can take into account questions such as “what targets were attacked?”

“What industries suffered?” “What, if any geopolitical factors played a role?” By answering these questions I can develop a more detailed understanding of why and how, or indeed why not, unanticipated attacks affect FDI inflow in Israel and Turkey.

Finally, for the sake of clarity I conduct the statistical analysis ceteris paribus meaning that wars, peace treaties, coups, and other potential economic disruptions are ignored. The impact of these issues will instead be discussed in the analysis section.

Results

The purpose of this study was to examine the effects of anticipated and unanticipated terrorist attacks on FDI net inflow into Israel and Turkey between the years 1975 and 2014. It further examined whether market size, GDP growth, inflation, and currency stability had any effects

(29)

29 on these results. For this reason, I conduct linear regression both with and without control variables for each independent variable.

This chapter is divided into three sections, first it will present descriptive statistics for my main dependent and independent variables. Second, correlations between anticipated and unanticipated terrorism are shown. Finally, the results of the linear regression model and the linear regression model with control variables are presented for each country respectively.

Descriptive Statistics

This section starts by presenting the descriptive statistics of the raw data used to create my dependent and independent variables. Following are descriptives of the actual variables used in the model.

Number of attacks, deaths and wounded in terrorist attacks Table 1

Range Minimum Maximum Mean SD Variance

Israel

Number of attacks 292 1 293 50.13 58.15 3381.66

Number of people killed 327 0 327 39.56 57.46 3301.48

Number of people wounded 1451 0 1451 170.27 252.89 63952.34

Turkey

Number of attacks 515 0 515 78.44 109.79 12054.75

Number of people killed 1247 0 1247 141.22 286.92 82325.40

Number of people wounded 778 0 778 135.38 170.10 28934.51

Note: n= 45

Table 1. Descriptive statistics showing the mean number of attacks, number of people killed, and number of people wounded in Israel and Turkey.

Descriptive statistics for independent variables Table 2

N Range Minimum Maximum Mean SD Variance

Israel

Spectrum variable attacks 40 20.93 -1.42 19.51 1.55 4.44 19.70

(30)

30

Spectrum variable deaths 40 9.08 -2.34 6.74 0.34 2.04 4.16

Spectrum variable wounded 40 5.92 -1.49 4.43 0.37 1.73 3.00

Binary variable attacks 40 1.00 0.00 1.00 0.25 0.44 0.19

Binary variable deaths 40 1.00 0.00 1.00 0.18 0.39 0.15

Binary variable wounded 40 1.00 0.00 1.00 0.23 0.42 0.18

Turkey

Spectrum variable attacks 40 15.83 -3.26 12.57 0.90 3.13 9.79

Spectrum variable deaths 40 22.13 -1.76 20.37 1.44 4.39 19.28

Spectrum variable wounded 40 12.35 -2.30 10.06 0.69 2.75 7.59

Binary variable attacks 40 1.00 0.00 1.00 0.33 0.47 0.23

Binary variable deaths 40 1.00 0.00 1.00 0.30 0.46 0.22

Binary variable wounded 40 1.00 0.00 1.00 0.25 0.44 0.19

Table 2 shows the descriptives of the independent variables used. Note that Israel has a higher mean spectrum variable whilst having a smaller mean for binary attacks. This tells us that years with unanticipated attacks in Israel had a larger number of unanticipated attacks than years with unanticipated attacks in Turkey.

Table 3 shows the distribution of anticipated and unanticipated attacks in Israel. Note that one unit of analysis is one year, thus 10 out of 40 years had unanticipated terrorist attacks. These numbers are applicable to both binary and Spectrum variable operationalisations.

Anticipated and unanticipated attacks Israel Table 3

Frequency Percent

Anticipated 30 75.00

Unanticipated 10 25.00

Total 40 100.00

Anticipated and unanticipated attacks Turkey Table 4

Frequency Percent

Anticipated 27 67.5

Unanticipated 13 32.5

Total 40 100.00

(31)

31 Table 4 shows the distribution of anticipated and unanticipated attacks in Turkey. Note that one unit of analysis is one year, thus 13 out of 40 years had unanticipated terrorist attacks. These numbers are applicable to both binary and Spectrum variable operationalisations.

Descriptive statistics for dependent variables Table 5

Table 5 shows the descriptive statistics of the dependent variable. Note that the mean change in Turkey is slightly larger than the mean change in Israel although by a relatively small margin. The maximum change is much larger than the minimum change for both countries indicating that when the opportunity seemed ripe investors were not hesitant to invest.

Correlations

Correlations Israel

N Range Minimum Maximum Mean

Std.

Deviation Variance Percentage change

FDI Turkey 40 519. 0 -91.2 427.77 44.12 16.15 102.17 10439.1

Percentage change

FDI Israel 44 445.0 -81.4 363.636 39 14.01 92.96 8642.41

Valid N (list wise) 40

Table 6

Spectrum variable attacks

spectrum variable deaths

spectrum variable wounded

Binary attacks

binary deaths

binary wounded Percentage

change FDI

Israel -.358** -0.28* -.372** -0.28* -0.30* -.365**

Change GDP

growth Israel -0.20 -0.10 -0.05 0.01 -0.04 0.04

Change market

size Israel -0.02 -0.10 -0.07 -0.09 -0.15 -0.06

Change official exchange rate

Israel -0.05 0.24 0.17 -0.01 0.30* 0.21

Change 0.01 -0.21 -0.10 0.03 -0.25 -0.20